一個具有竄改定位能力的可逆式影像驗證機制

A Reversible Image Authentication Scheme with the Capacity of Tamper Localization

鄭君偉(Chun-Wei Cheng)

指導教授 ： 黃育銘

國立暨南國際大學/科技學院/資訊工程學系/碩士(2007年)

摘要

現今的數位影像處理軟體功能日益強大，操作介面也更加簡便，尤其是數位影像易於編輯與複製的特性，今天當接收者要了解所收到影像是否有遭到別人惡意的竄改，而失去影像內容原來的意義，此時就需要影像驗證的輔助。在傳統的影像驗證有著一項缺點，也就是當嵌入驗證訊息到影像當中時，會造成影像的些許失真，這一點在某些領域是不被允許的，如:醫療、軍事、法律等應用。因此，可逆式影像驗證架構便被提出來，如果接收者已對影像作完驗證的動作，確認沒有遭到竄改，便可以將嵌入的浮水印完全的移除，讓影像回復到原始的狀態，沒有任何的差異，這類文章在2001年最早被提出，但在此之後的文章都只有簡單的驗證能力，卻缺乏了影像驗證最重要的功能-竄改定位能力，而在2006年Celik et al.提出了一個新的可逆式影像驗證架構，並且包含了竄改定位的能力。
本文基於Celik et al.的架構上，改變了其原始訊息做可逆式資料嵌入的部份，由於原方法雖可達到不錯的影像品質，但會隨每張影像的不同，而造成影像品質會有高低落差，而這裡參考了Lee et al.的文章，利用其穩定的嵌入影像品質，並改善了其像素溢位的問題，且增加了竄改定位的功能，藉以應用在新的可逆式影像驗證的架構中，同時也增加了具唯一性的影像代碼，以增加安全性，使本系統同樣能夠防止向量量化攻擊，又同時可達到可逆式影像驗證功能及正確的竄改定位能力，並提供穩定且不錯的嵌入影像品質，只要嵌入的訊息量固定，便可以計算出影像品質的最低門檻值。

關鍵字

影像驗證 ； 可逆式影像驗證 ； 可逆式資料嵌入 ； 竄改定位 ； 向量量化攻擊

並列摘要

Digital image processing software is now readily available. Indeed, due to the popularity of internet, digital image can be duplicated, manipulated and transformed very easily. Image authentication is a new rising technique to let the receiver know whether the received image was tampered or not. In the traditional watermarking-based image authentication scheme, the host image usually needs to be modified to accommodate the authentication information and it will induce some distortion in image fidelity. In some applications, such as medical, military, and legal imaging applications, even imperceptible distortion is not permissible. Therefore, the reversible image authentication schemes have been widely explored recently. After verifying the integrity of the watermarked image, then the watermark can be removed thoroughly from the watermarked image to obtain the exact copy of the original un-watermarked image by using the so-called reversible image authentication technique.
Before 2006, in the literature, none of the existing lossless authentication methods offer the capability of tamper localization which is one of the major functions of authentication scheme. In 2006, Celik et al. first proposed a new reversible image authentication framework that encompasses the ability of tamper localization.
In this thesis, based on the Celik`s framework, we implement the reversible image authentication scheme proposed by Lee. We first resolve the pixel overflow/underflow problem inherent in original Lee’s method through an unique image index, which can also provide good security and capability against vector quantization attack. Experimental results show that the modified scheme can own more stable watermarked image quality. While the length of the embedding message is fixed, the lower bound of the PSNR value of the watermarked image can be easily obtained.

並列關鍵字

Image authentication ； reversible image authentication ； reversible data embedding ； tamper localization ； vector quantization attack

參考文獻

[1]C. C. Chang, Y. S. Hu, and T. C. Lu, “A watermarking-based image ownership and tampering authentication scheme,” Pattern Recognition Letters 27, pp. 439-446, Oct. 2005.

[2]P. L. Lin, C. K. Hsieh, and P. W. Huang, “A hierarchical digital watermarking method for image tamper detection and recovery,” Pattern Recognition 38, Feb. 2005, pp. 2519-2529.

[3]M. U. Celik, G. Sharma, A. M. Tekalp, and E. Saber, “Hierarchical watermarking for secure image authentication with localization,” IEEE Trans. on Image Processing, vol. 11, no. 6, pp. 585-595, 2002.

[4]Y. L. Tang, and C. T. Chen, “Image authentication using relation measures of wavelet coefficients,” presentenced at the Proc. IEEE Int. Conf. on e-Technology, e-Commerce and e-Service, pp. 541-545, Mar. 2004.

[5]C. Y. Lin, and S. F. Chang, “A robust image authentication method distinguishing JPEG compression from malicious manipulation,” IEEE Trans. on Circuit and system of video technology, vol. 11, no. 2, Feb. 2001.